基于改进遗传算法的贝叶斯入侵检测研究

发布时间：2018-01-02 07:00

本文关键词：基于改进遗传算法的贝叶斯入侵检测研究　出处：《湖北工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：入侵检测系统作为防火墙的补充,可以通过主动防御的方式提前预判网络环境中存在的安全隐患并及时做出响应,协助防火墙保障网络安全。入侵检测系统的核心技术在于检测算法。高效、稳定的检测算法能通过较少的特征数据对网络中的连接进行精准识别,达到提前预警的目的。本文针对入侵检测系统中的检测算法进行研究,在学习和研究同类算法的基础上展开讨论。全文主要工作如下:(1)介绍了入侵检测系统的技术背景,国内外研究现状及本文的研究目的和方法。介绍了入侵检测KDD99数据集,并对数据预处理方法进行详细说明,将贝叶斯算法用于KDD99数据集的分类,并对分类结果做详细分析。(2)针对贝叶斯分类结果和KDD99数据集高维度特性,提出将遗传算法用于数据集进行特征选择,并将选出的特征子集用于贝叶斯分类,以求获得更好的分类性能。在选择最优个体时,由于适应度函数是多元函数,本文针对精英保存策略的不足,提出一种个体评优的方法。在该方法中,适应度函数值不再是评估个体优劣的唯一标准,而是综合考虑影响个体最优的各种因素,并将影响因素按照重要程度赋予不同影响因子,在参照适应度函数值为标准的基础上提出了个体评优的新标准。交叉操纵过程中,本文引入基因相似度的概念,对个体基因相似程度进行比较,并阻止基因相似度超过设定阈值的近亲个体进行交叉,以提高种群个体多样性。在变异过程中,为加快收敛速度,本文采用动态非线性变异算子对种群个体进行操作,这种操作方式在加快收敛速度的同时还能保证种群进化的随机性,更符合遗传学理论。(3)将本文所提的改进遗传算法用于KDD99数据集进行特征选择,并将选择出的特征子集采用贝叶斯算法进行分类。为验证本文所提方法的有效性,共设计三组不同对比实验,实验结果表明:采用本文改进的遗传算法能获得更低的数据维度且能提高检测率。
[Abstract]:The intrusion detection system as a complement of firewall, security risks can be through the active defense way to predict in advance the network present in the environment and respond in time, assist the firewall to protect network security. Intrusion detection system is the core technology of the detection algorithm. Efficient detection algorithm can be stable for precise identification of connections on the network through the characteristics of less data the reach the aim of early warning. Detection of the intrusion detection system in this paper, the discussion on the basis of the learning and research of similar algorithms. The main works are as follows: (1) introduces the technical background of the intrusion detection system, the research purpose and methods and the status of research at home and abroad. This paper introduces the intrusion detection of the KDD99 data set, and a detailed description of the method of data preprocessing, the Bias algorithm for classification of KDD99 data sets, and every kind of results For a detailed analysis. (2) according to the Bias classification results and KDD99 data sets of high dimension characteristics, proposed a genetic algorithm for the data set for feature selection and feature subset for classification of Bias, in order to obtain a better classification performance. In the selection of the best individual, the fitness function is multi function, aiming at the elite preservation methods, proposes a method for individual evaluation. In this method, the fitness function value is no longer the sole criterion for individual merits, but considering the impact of various factors of individual optimal, and the factors in accordance with the importance of different factors, the reference value of fitness function is based on the standard put forward a new standard of individual appraised. Cross operation process, this paper introduces a concept of genetic similarity, the similarity degree of individual genes were compared, and stop gene similarity Exceeds the preset threshold relative cross individual, in order to improve the population diversity. In the process of mutation, in order to speed up the convergence, this paper uses the nonlinear dynamic mutation operator to operate on the individual, this mode of operation can accelerate the convergence speed while ensuring the randomness of population evolution, more in line with the theory of genetics (3) will be. The improved genetic algorithm proposed in this paper for the KDD99 data set for feature selection and feature subset selected using Bayesian classification algorithm. In order to verify the validity of the method proposed in this paper, a total design of three different groups of contrast experiments, experimental results show that the improved genetic algorithm used in this paper can obtain the data of lower dimension and to improve the detection rate.

【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.08;TP18

【相似文献】